Estimating the Long-Run Relationship between Market and Accounting Values Using Cross Section Analysis: The Need for Dynamic Firm-Level Modeling in Capital Market Research

Abstract

Average parameter estimates from dynamic models for 30 large US firms over the period 1955-2004 are compared to the average parameter estimates from 50 annual cross sections. Results from this sample show statistically well specified models are multiplicative in form, not additive. Dynamic characteristics of model variables complicate interpretation of cross section parameters as estimates of the long-run. Certain types of heterogeneity in parameters between firms may prevent recovery of average long-run effects from cross section models. Cross-section 'levels' regressions may yield valid, reasonably accurate estimates of long-run effects. Cross-section returns regressions typically implemented are unlikely to do so.

Keywords: Cross section modeling, misspecification, heterogeneity, dynamic modeling, long run effect, fundamentals

JEL Classification: C21, C22, C51, C52, M41